The Relationship Between Cognition and Cerebrovascular Reactivity: Implications for Task-Based fMRI - Frontiers
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
REVIEW
published: 12 April 2021
doi: 10.3389/fphy.2021.645249
The Relationship Between Cognition
and Cerebrovascular Reactivity:
Implications for Task-Based fMRI
Rebecca J. Williams 1*, M. Ethan MacDonald 1,2 , Erin L. Mazerolle 3 and G. Bruce Pike 1
1
Department of Radiology, Clinical Neuroscience and Hotchkiss Brain Institute, University of Calgary, Calgary, AB, Canada,
2
Department of Electrical and Software Engineering, University of Calgary, Calgary, AB, Canada, 3 Department of Psychology,
St. Francis Xavier University, Antigonish, NS, Canada
Elucidating the brain regions and networks associated with cognitive processes has
been the mainstay of task-based fMRI, under the assumption that BOLD signals are
uncompromised by vascular function. This is despite the plethora of research highlighting
BOLD modulations due to vascular changes induced by disease, drugs, and aging. On
the other hand, BOLD fMRI-based assessment of cerebrovascular reactivity (CVR) is
often used as an indicator of the brain’s vascular health and has been shown to be
Edited by:
strongly associated with cognitive function. This review paper considers the relationship
James Duffin,
University of Toronto, Canada between BOLD-based assessments of CVR, cognition and task-based fMRI. How the
Reviewed by: BOLD response reflects both CVR and neural activity, and how findings of altered CVR
Jean Chen, in disease and in normal physiology are associated with cognition and BOLD signal
University of Toronto, Canada
Tommaso Gili, changes are discussed. These are pertinent considerations for fMRI applications aiming
IMT School for Advanced Studies to understand the biological basis of cognition. Therefore, a discussion of how the
Lucca, Italy
acquisition of BOLD-based CVR can enhance our ability to map human brain function,
*Correspondence:
with limitations and potential future directions, is presented.
Rebecca J. Williams
rjwillia@ucalgary.ca Keywords: functional magnetic resonance imaging, cerebrovascular reactivity, cognition, aging, caffeine, multiple
orcid.org/0000-0002-8949-1197 sclerois, Moyamoya, small vessel disease
Specialty section:
This article was submitted to INTRODUCTION
Medical Physics and Imaging,
a section of the journal Cerebrovascular reactivity (CVR) reflects the ability of the cerebral blood vessels to respond to
Frontiers in Physics
a vasoactive stimulus and is primarily sensitive to arterial compliance [1, 2]. Blood oxygenation
Received: 23 December 2020 level-dependent (BOLD) functional MRI is an effective technique for measuring CVR. The BOLD
Accepted: 11 March 2021 signal originates from changes in the local concentration of deoxyhemoglobin (dHb), which
Published: 12 April 2021
primarily occurs in the post-arteriolar component of the cerebral vasculature. The change in the
Citation: local concentration of dHb is predominantly driven by increases in cerebral blood flow (CBF),
Williams RJ, MacDonald ME,
which reduces venous dHb via a dilution effect and thus increases the BOLD signal [3, 4]. In
Mazerolle EL and Pike GB (2021) The
Relationship Between Cognition and
conventional task-based or resting-state BOLD fMRI, increases in the cerebral metabolic rate of
Cerebrovascular Reactivity: oxygen (CMRO2 ) and cerebral blood volume (CBV) associated with increased neuronal activity
Implications for Task-Based fMRI. both act to increase dHb and hence attenuate the CBF dominated BOLD signal increase. With a
Front. Phys. 9:645249. direct vasodilator stimulus used for CVR mapping, such as carbon dioxide (CO2 ), CMRO2 changes
doi: 10.3389/fphy.2021.645249 are assumed to be negligible and a robust positive BOLD signal is measurable that primarily reflects
Frontiers in Physics | www.frontiersin.org 1 April 2021 | Volume 9 | Article 645249
Williams et al. CVR, Cognition and Task fMRI
the increase in CBF [5, 6]. CVR can also be measured using CVR AND BOLD SIGNAL
other MRI techniques including those sensitized exclusively to
CBF (e.g., arterial spin labeling), macrovascular blood flow (e.g., Cerebral blood flow (CBF) is normally controlled by regulatory
phase contrast flow imaging), or CBV [1, 7–9]. In addition, there mechanisms including cerebral autoregulation, the partial
are non-MRI techniques available for measuring CVR, such as pressure of arterial blood gases, and neurovascular coupling
transcranial doppler, which measures blood flow velocity within (NVC) [21]. NVC is discussed in detail in section Neural Activity
the middle cerebral artery [10], and [15 O] positron emission and BOLD Signal. Cerebral autoregulation keeps blood flow
tomography [11, 12]. Yet BOLD fMRI is the more broadly used consistent with changing cerebral pressure. The Hagen-Poiseuille
modality for CVR mapping in clinical and research settings equation describes the changes in cerebrovascular resistance as:
owing to its sensitivity to vascular properties, combined with its
safety, availability, higher spatial resolution, and reproducibility πPr4
F=
[13]. The sensitivity and spatial resolution of BOLD fMRI 8ηl
also explain its popularity as a cognitive neuroscience tool
for mapping neural activity. Neural activity increases oxygen Where blood flow (F) is proportional to cerebral pressure (P)
metabolism; however, BOLD signal increases occur in the context and vessel radius (r) and inversely proportional to fluid viscosity
of increased CMRO2 due to coupled, but proportionally larger, (η) and vessel length (l). The purpose of cerebral autoregulation
CBF increases [14–16]. is to ensure continued blood supply to the brain across broad
While CVR primarily assesses vascular compliance, it is variations in blood pressure, which is achieved by manipulating
also evident that CVR is linked to cognitive functioning. vascular resistance. Changes in cerebrovascular resistance are
Much of this evidence has emerged from research investigating also observed with changes in partial pressure of arterial CO2
pathological conditions such as vascular disease and dementia, (PaCO2 ) [22]. The relationship between PaCO2 and CBF is
where both regulation of cerebral vasculature and cognitive sigmoidal; when PaCO2 is between ∼20 and 60 mm Hg, there
function are affected [17–19]. There are differences in the is an increase in CBF of ∼3–4% for every 1 mm Hg increase in
underlying mechanisms leading to the resultant hyperemia in PaCO2 . The CBF response is diminished above this PaCO2 range
CVR and task-based BOLD activation, however both rely on [21, 23, 24]. The increased CBF in response to PaCO2 decreases
vessel contraction and dilation. Furthermore, the BOLD response the concentration of dHb via dilution and robustly increases
to hypercapnia has shown to account for a large amount of the BOLD signal, allowing for significant signal detection; there
the task-induced BOLD variability [20]. Therefore, the aim of is a relatively large BOLD signal difference (∼1–2%) between
this paper is to review the link between CVR and cognitive isocapnic and hypercapnic states, when a moderate level of
processes, and to highlight the importance of acquiring CVR hypercapnia is induced (∼5–8 mm Hg above baseline levels).
data to accompany task-based fMRI data when investigating the There are numerous mechanisms leading to vasodilation.
biological underpinnings of cognition. Increased CO2 results in increased carbonic acid in the blood,
The first two sections of this review outline how CVR that leads to hyperpolarization of the smooth muscle cells and
and neural activity are mapped using BOLD fMRI. Initially, vasodilation [2]. CO2 in the blood lowers the pH and changes
an overview of how changes in arterial CO2 are coupled to the muscle tonus of the vessel wall [25]. An increase in CBF
modulations in CBF are presented. The following section is is observed following this reduction in pH, which functions to
focussed on task-based fMRI and discusses the mechanisms washout the excess CO2 and regulate pH levels. The increased
underpinning neurovascular coupling. Both of these sections CO2 in the cerebral spinal fluid (CSF) from hypercapnia is known
emphasize recent findings. In sections Neurological Conditions to result in vascular relaxation [26]. Central chemoreceptors
and Normal Physiology, focus is shifted to example situations located in regions throughout the brainstem, cerebellum, and
where both CVR and cognitive functioning are altered in the hypothalamus, which are sensitive to CSF pH, modulate
study population. We demonstrate that modulations to task- respiratory rate to regulate CO2 levels [27]. Indeed, increasing the
based BOLD activation are also often found in such cases, fraction of inspired CO2 stimulates the breathing rate to increase,
and provide the motives for collection of CVR data to aid the in order to achieve CO2 elimination [28].
interpretation of fMRI activation maps. Neurological diseases End-tidal partial pressure of CO2 (PET CO2 ) is a close proxy
hallmarked by vascular and microstructural variations are measure of PaCO2 and is therefore used in CVR studies. The
discussed in detail. However, vascular and cognitive alterations total volume of exhaled air is comprised of CO2 within the alveoli
are also observed in normal physiology, such with caffeine and the physiological dead space. Physiological dead space is
consumption and aging, and this is discussed in section inhaled air that does not participate in gas exchange and includes
Normal Physiology. Finally, it has become increasingly clear non-perfused air within the alveolar and anatomical dead space.
that the integration of multiple imaging modalities improves Anatomical dead space is the air within the nose, trachea and
our ability to map the human brain. Techniques for correcting bronchi that make up the conducting airways. The function of
BOLD-based assessments of brain activity using CVR (i.e., the conducting airways is to channel inhaled air to the respiratory
hypercapnic normalization) and the advantages and pitfalls of zone where gas exchange occurs; that is, the alveolar surface
this analytical approach, are discussed in section Integration [29, 30]. In healthy adults, the alveoli dead space is negligible
of BOLD-Based Assessments of CVR and Neural Activity: and anatomical dead space comprises all of the non-perfused
Hypercapnic Normalization. air. However, excessive air within the alveoli that is not perfused
Frontiers in Physics | www.frontiersin.org 2 April 2021 | Volume 9 | Article 645249
Williams et al. CVR, Cognition and Task fMRI
increases the alveoli component of dead space [30]. This excess directly detect changes in neural activity, and therefore may be
air within the alveoli typically indicates less efficient CO2 removal more involved in NVC than previously thought [50–52].
and is associated with pulmonary diseases. It will result in a Neural functioning requires a balance of excitation and
slightly lower (∼ ≤5 mm Hg) PET CO2 relative to PaCO2 [31], inhibition, hence recent work has also focused on understanding
however, this does not preclude PET CO2 from being a robust how inhibitory neurons generate vascular changes. There
indicator of PaCO2 , and is used in the determination of CVR, is evidence from electrophysiology studies demonstrating
calculated as 1%BOLD/1PET CO2 (mm Hg) [2]. that gamma power local field potentials, driven by γ -
Elevated PaCO2 results in a robust vasodilatory response and aminobutyric acid (GABA) interneurons, are correlated with
CBF increase, and therefore, experimental manipulations that hemodynamic changes [53]. GABAergic interneurons are diverse
increase PET CO2 are used to measure CVR. Breath-holding is in morphology and physiology, with subtypes characterized by
one such approach that is low-cost and low-risk. A benefit to their gene expression and cell innervation [54]. They play an
the breath-hold approach is that it is more accessible than gas essential role in cortical function by inhibiting excitatory or
inhalation because it requires less equipment. The monitoring of inhibitory post-synaptic neurons, and while connections remain
PET CO2 is highly desirable however, because of variability in the mostly focal, some GABAergic interneurons extend to the
participant breath-hold performance that can be accounted for if vasculature [55].
measured PET CO2 is included as a statistical regressor in CVR Significant progress has been made in understanding the
quantification [32]. Comparisons of breath-hold paradigms to unique role of GABAergic cells in NVC using optogenetic
CO2 inhalation methods have shown consistent CVR results [33, stimulation [56, 57]. Selective photo-stimulation of inhibitory
34], supporting this method as a reliable technique for calculating cells can be achieved using a mouse model expressing
CVR. Inhaling gas mixtures containing increased concentrations channelrhodopsin-2 (ChR2) in GABAergic neurons. ChR2 is
of CO2 is controllable and more precise than breath holding, a light-sensitive cation channel that can induce neuronal
and therefore a desirable approach to increasing PaCO2 . This depolarization [58]. A study by Vazquez et al. [59] used
is achieved using MRI-compatible breathing circuits that allow a ChR2 mouse model and targeted photo-stimulation to
for the administration of gas mixtures and the precise recording selectively activate GABAergic interneurons. Blood flow, volume,
of PET CO2 [35]. There are different types of breathing circuits and oxygenation increases were observed following photo-
available; further information on these can be found in references stimulation of GABAergic interneurons. Stimulation of certain
[2] and [36]. GABAergic neurons has also resulted in decreases in blood
volume in adjacent cortical tissue [60]. This decrease in blood
volume is consistent with suggestions that negative BOLD fMRI
NEURAL ACTIVITY AND BOLD SIGNAL may reflect neural inhibition [61–63]. Indeed, there is increasing
evidence supporting the role of inhibitory neurons in driving
Task-based BOLD fMRI reflects neural activity indirectly through NVC and shaping the BOLD response [64]. Some of the work
NVC [15, 37–40]. NVC is the tightly linked relationship between investigating the negative BOLD response is discussed in greater
regional neural activity and changes in blood flow [39]. It is detail the next section.
independent of perfusion pressure, and results from the close
communications between neurons, glia, and arterioles. The
functional hyperemia coupled to neural activity changes ensures Hemodynamic Responses to Neural
that activated neural cells have a constant supply of oxygen and Activity and CO2
glucose [41]. Arteries commence dilation within hundreds of The traditional positive BOLD response (PBR) to neural
milliseconds of neuronal activity, and return to baseline within activation is modeled in task-based fMRI using a hemodynamic
seconds of stimulus termination [14]. The venous side is slower response function (HRF). The HRF in its canonical form is
to dilate, taking tens of seconds, and appears to have smaller typically described using the sum of two gamma functions. In
blood volume changes than the arterial side [42]. There has gray matter, the time-to-peak is ∼6 s from the onset of neural
been a recent wealth of research focussed on describing the roles activity, however significant spatial heterogeneity in the shape
of specific cell types and their signaling resulting in arteriole and timing parameters of HRF has been observed in healthy
dilation [43]. For instance, there is strong evidence for the role brains [65, 66] and in different clinical populations [67, 68].
of glutamatergic cells in signaling the commencement of arterial Characterizing temporal HRF parameters may reveal important
dilation. Blood flow increases in conjunction with excitatory physiological information not evident from the amplitude of
pyramidal cell activity have been demonstrated in rodent models the response alone, which is the main parameter of interest
[44, 45]. CBF, and to a lesser extent CBV, have been shown to when calculating fMRI activation maps [68, 69]. HRF parameters
correlate with post-synaptic local field potentials but presynaptic such as time-to-peak and full-width at half-maximum are
activity does not appear to trigger hemodynamic changes [46]. influenced by blood flow and NVC, may be important for probing
Astrocytes are important in maintaining basal blood flow [47] information regarding neuronal duration [70].
and have been implicated in NVC due to their close proximity Focus has also been placed on understanding negative
to both the microvasculature and neurons [48], but how they BOLD responses (NBRs) in task-based fMRI. These responses
contribute to NVC remains the topic of debate [47, 49]. An are identified as stimulus-induced decreases in BOLD signal
intriguing hypothesis is that endothelial cells in blood vessels can intensity relative to baseline, or inverted PBRs. Different
Frontiers in Physics | www.frontiersin.org 3 April 2021 | Volume 9 | Article 645249Williams et al. CVR, Cognition and Task fMRI
hypotheses have been suggested to explain the underlying are microstructurally-driven. Therefore, Multiple Sclerosis is
mechanisms of the NBR. One such suggestion is centered on it discussed in section Multiple Sclerosis.
having purely vascular origins, emerged from research indicating
that negative BOLD changes may result from a redistribution Cerebral Small Vessel Disease
(steal) of CBF to adjacent, active cortical regions [71, 72]. Other Cerebral small vessel disease (SVD) refers to a collection
research has provided conflicting evidence; for instance, NBRs of abnormalities affecting the small blood vessels, including
have been identified in brain regions distant from simultaneously the capillaries, arterioles and venules, in the gray and white
occurring PBRs, with these NBR regions having distinct arterial matter [86]. SVD is strongly associated with advanced age, and
territories [61, 73]. Decreases in neuronal activity accompanying has been recognized as a common cause of stroke [87, 88]
NBRs have suggested a neural origin, with much of the literature and dementia [89]. Indeed, SVD is tightly interwoven with
supporting this [62, 63, 74–77]. Influential findings from AD. Due to the contribution of SVD to AD pathogenesis,
Shmuel et al. [63] demonstrated significant correlations between the distinction between the two is increasingly blurred [90].
NBRs and decreases in local field potentials and multiunit Imaging features of SVD are heterogenous and can present
activity, measured directly from the primary visual cortex of as white matter hyperintensities (WMH), lacunar infarcts,
monkeys. This has been verified in human work using combined microbleeds, perivascular spaces, and atrophy [88, 91]. SVD
electroencephalography (EEG) fMRI, where negative CBF and types can demonstrate regional preference, for instance, cerebral
BOLD changes in NBR regions demonstrate concomitant EEG amyloid angiopathy (CAA) pathology disproportionately affects
changes [75, 78]. Although there has been progress made in the occipital cortex. CAA is characterized by the accumulation
the attempt to understand the response properties of the NBR of amyloid-β predominantly in the arterial walls of the
[78, 79], further work elucidating how and when NBRs occur and leptomeningeal space and cerebral cortex [92, 93]. SVD types
the associated biophysical changes is essential for the accurate can also be distinguished based on etiology, with sporadic
interpretation of this signal. appearances being distinct from those with known genetic
Analyses of BOLD responses to vasoactive stimuli have shown etiologies, such as Cerebral Autosomal Dominant Arteriopathy
that temporal parameters may reflect important aspects of with Subcortical Infarcts and Leukoencephalopathy (CADASIL).
vascular function in CVR. Typically, CVR mapping considers Endothelial dysfunction has been suggested as contributing
only the amplitude of the BOLD response to the vasoactive to the pathogenesis of SVD, resulting in stiff vessels with low
stimulus; however, accounting for temporal variations across reactivity [94, 95]. Concordant results have been reported in
the brain may improve CVR estimation accuracy. Identifying studies implementing CO2 challenges to measure CVR in SVD
temporal variation may also provide novel physiological patients. Lower gray matter CVR has been demonstrated in
information. For instance, a slower response may reflect slower SVD patients [96], and in patients with AD and mild cognitive
vessel dilation due to vascular pathology. The speed of the impairment [97], compared to healthy controls. CVR reductions
BOLD response to changes in PET CO2 has shown to reflect may show spatial heterogeneity, with AD patients showing
pathophysiology in patients with steno-occlusive disease [80]. reduced anterior (prefrontal, insular, anterior cingulate) CVR
Slower BOLD responses to CO2 have also been demonstrated in [98]. This study also found that the brain regions with reduced
patients with mild cognitive impairment and Alzheimer’s Disease CVR did not overlap with regions showing reduced baseline
(AD) [17]. Analytic approaches to account for temporal delays CBF in AD patients, owing to vasoreactivity being sensitive to
have been put forth [81–83] and shown to improve CVR mapping different pathology than CBF. Lower CVR in both gray and white
in patients with blood flow abnormalities [84]. matter in SVD is positively associated with WMH volume [99],
and increases in WMH over time in patients with CADASIL
[100]. A systematic literature review was performed by Blair
NEUROLOGICAL CONDITIONS et al. [95], who suggested that the relationship between CVR and
WMH is unclear due to inconsistent findings [95]. The authors of
Abnormal NVC, or even complete uncoupling, can occur when this review further suggest that differences in imaging protocols
the cascade of events from neuronal activity to functional might contribute to these inconsistencies. Another limitation is
hyperemia is disrupted. This is known to occur in clinical the lack of research incorporating temporal information into
conditions such as in brain tumors and cerebrovascular disease, CVR analyses, which has shown to improve CVR estimation
and CVR mapping may be effective in improving task-based and provide important physiological information in patients
fMRI interpretation in these instances [85]. The utility of CVR with SVD [101]. Furthermore, more longitudinal research is
may extend beyond mapping vascular changes to also inform warranted, as regions with reduced CVR have shown to precede
our understanding of cognition. In this section, this argument the appearance of WMH [102].
is made by placing focus on example neurological conditions Cognitive impairment is a common occurrence in patients
where CVR, cognitive and task-based fMRI alterations occur. with SVD. The severity and type of cognitive impairment is
The first two diseases discussed, cerebral small vessel disease and highly variable across patients, which may reflect the differences
Moyamoya Disease, are characterized by vascular abnormalities. in pathology load and location [103–106]. Cognitive domains
Such vascular-driven diseases are obvious candidates to benefit frequently impaired in SVD include processing speed and
from CVR mapping. However, the utility of CVR may executive function [107], however patients tend to show impaired
extend beyond diseases with vascular etiologies, to those that cognitive abilities across a broad range of domains [108]. This
Frontiers in Physics | www.frontiersin.org 4 April 2021 | Volume 9 | Article 645249Williams et al. CVR, Cognition and Task fMRI
is demonstrated by Table 1, which highlights the range of [123] show promising results for CVR in patients with MMD.
cognitive impairment observed in SVD. This table summarizes These findings demonstrate the utility of CVR for assessing
findings from example research studies implementing behavioral which brain regions are afflicted by changes in cerebral perfusion
assessments to test different cognitive domains in these patients. in MMD, and for tracking post-intervention reorganization.
The studies reported in Table 1 were found using a PubMed Vascular steal is commonly observed in CVR maps in patients
search, and search terms included medical subject headings with MMD. This phenomenon is observed as negative CVR
(MeSH) “cerebral small vessel disease,” and each of the within brain regions affected by arterial narrowing when a
assessable cognitive domains [115]. The searched domains were vasoactive stimulus such as CO2 is applied [124]. Negative CVR
“perception,” “motor skills,” “attention,” “memory,” “executive can be observed in an example patient in Figure 1A below.
functioning,” “processing speed,” “language,” and “verbal skills.” NBRs in the context of CVR are attributed to the permanently
When selecting articles for Table 1, preference was given to vasodilated state of the affected vessels, which is necessary
recently published literature with an age and gender-matched to ensure sufficient perfusion in the resting baseline state.
control group. When PaCO2 increases, these vessels cannot decrease vascular
Investigating potential imaging markers of SVD, including its resistance any further to meet this change in demand. Therefore,
pathogenesis and ongoing physiological changes, is an essential blood flow is diverted from these regions to brain regions
research endeavor [18]. Task-based BOLD fMRI changes in with intact vasodilatory response [126]. Vascular steal in MMD
SVD have shown to reflect CVR and cognitive findings, and has been associated with increased risk of stroke [127, 128],
may potentially provide imaging markers of disease processes. cortical thinning [129], and increased diffusion in white matter
Sensory tasks targeting activity in the visual and motor cortices [130]. However, pre-operative appearance of negative CVR has
report reduced BOLD signal amplitude in the visual cortex in been associated with improved surgical outcome [131]. Negative
patients with CAA [116]. Analysis of the temporal characteristics CVR has shown to resolve following revascularization surgery,
of these BOLD signal reveals wider and delayed responses in representing normalized cerebral perfusion [126, 132].
CAA patients compared to aged-matched healthy controls [68]. Cognitive deficits have been described as a common symptom
Longitudinal research has shown that BOLD responses in the in patients with MMD, with ∼30% of patients demonstrating
occipital cortex of CAA patients decrease over time [117]. These cognitive impairment [133]. Cognitive studies in pre-surgical
BOLD signal changes may be specific to CAA due to pathology MMD are summarized in Table 2, where the same search
largely affecting the occipital cortex; no such BOLD activation strategy outlined for SVD and Table 1 in section Cerebral
reductions to visual and motor tasks were observed in CADASIL Small Vessel Disease above was applied. Cognitive impairment
patients [118]. SVD patients have also demonstrated alterations in patients with MMD may be attributed to hypoperfusion
in BOLD activation to cognitive fMRI tasks. Reductions in BOLD [138], as chronic ischemia may cause reduced axon density
signal amplitudes and spatial extent of activation to an attention [139] and decrease cognitive network communication efficiency.
task have been reported in patients with SVD [119]. This is While task-based fMRI can be used to clarify neural networks
supported by resting-state fMRI showing reduced frontoparietal associated with certain cognitive tasks, regional ischemia and
connectivity in these patients [120]. Disease processes disrupting perfusion alterations in MMD complicate the ability to identify
neural networks may cause changes in neural activity, reflected by neural activity through a BOLD response. Mazerolle et al.
the cognitive and BOLD data. However, vascular abnormalities, [140] showed this in pre-operative MMD patients when visual-
as reflected by CVR findings, will also affect task-based BOLD motor tasks were performed. When patients performed a
signal in these patients and present a challenge when interpreting simultaneous visual-motor task, BOLD activation in the motor
BOLD activation maps. When implementing fMRI to investigate cortex of the afflicted hemisphere was not detectable, although
SVD imaging markers, CVR may be an important addition to the prominent activation of the visual cortex was observed. However,
imaging protocol by providing insight into vascular changes and when the motor task was performed in isolation (i.e., without
guide interpretation of fMRI activation. concurrent visual stimulation), robust BOLD activation was
detected in the afflicted hemisphere. Therefore, the activation of
Moyamoya Disease the visual cortex, which had intact CVR, resulted in vascular steal
Moyamoya disease (MMD) is an idiopathic vasculopathy, from the MMD-afflicted motor cortex with impaired reactivity.
characterized by reduced regional cerebral perfusion due to Figures 1B,C highlights altered task-based fMRI activation maps
arterial occlusion or stenosis. From the branches of these during a concurrent visual-motor task. These results support the
narrowed arteries, collateral vascular networks form which necessity of CVR when fMRI is performed in patients with MMD.
radiologically appear as a “puff of smoke” (which gave
the Japanese name Moyamoya) [121]. Patients with MMD Multiple Sclerosis
often present with headaches, transient ischemic attacks, Multiple sclerosis (MS) affects the central nervous system and
and intracranial hemorrhage. Intervention typically involves is characterized by focal regions of demyelination, gliosis, and
revascularization surgery. This improves cerebral perfusion to axonal degeneration. The majority of patients demonstrate
the cerebral vascular territories afflicted, and has shown to have a relapsing-remitting (RRMS) form at onset, hallmarked by
good outcome regarding mortality and quality of life [122]. discrete episodes of neurological impairment followed by
Comparisons of BOLD CVR to gold-standard CBF mapping periods of remission [141]. Gradual neurodegeneration leads
using [15 O] positron emission tomography [11] and angiography to secondary progressive MS (SPMS), where remission periods
Frontiers in Physics | www.frontiersin.org 5 April 2021 | Volume 9 | Article 645249Williams et al. CVR, Cognition and Task fMRI
TABLE 1 | Summary of research studies highlighting the cognitive domains impaired in cerebral small vessel disease.
References Study population(s) N Mean age Cognitive domain(s) Tests administered Main findings
(N Female) assessed
Dey et al. [109] SVD 23 (12F) 71 ± 5.5 Attention, Executive Neuropsychology test Groups differed in subjective SC
functioning battery (P = 0.008), but not
neuropsychological tests.
Healthy older adults 23 (14F) 69.3 ± 4.6 Subjective cognitive Cognitive Failures
(HOA) complaints (SC) Questionnaire; Dysexecutive
Questionnaire
Su et al. [110] SVD 68 (27F) 66.2 ± 8.1 Motor skills Lower and upper SVD performed worse than HOA
extremities- behavioral tests on lower extremities
tests (P < 0.01).
HOA 702 (474F) 56.3 ± 9.0
Metoki et al. [111] SVD 106 (36F) 69.5 ± 9.6 General cognitive Mini-Mental State SVD group had significantly
impairment Examination (MMSE) lower MMSE scores (P < 0.001).
HOA 35 (13F) 69.6 ± 9.3 Working memory (WM) Digit Span Test of WMS-III† SVD group had significantly
lower WM scores (P < 0.001).
Long-term memory (LTM) Immediate and delayed SVD group had significantly
recall subtests of WMS-III† lower LTM scores (P < 0.001).
Cao et al. [112] SVD with vascular 32 (9F) 67.5 ± 9.6 All cognitive domains Neuropsychology test VaMCI performed worse than
mild cognitive battery NCI on all domains (P < 0.01)
impairment (VaMCI) except language.
SVD with no cognitive 23 (6F) 67.4 ± 8.5
impairment (NCI)
Lawrence et al. SVD (over 3 years) 98 (33F) 69 ± 9.9 All cognitive domains Neuropsychology test Executive functioning was the
[113] battery only cognitive domain to
significantly decline over 3 years.
Herbert et al. SVD 45 (20F) 69.7 ± 8.2 Verbal fluency (phonemic COWAT‡ ; Animals and SVD group performed worse on
[114] and semantic) Tools Tests both verbal fluency tests
compared to HOA (P < 0.001).
Alzheimer’s Disease 24 (12F) 74.5 ± 6.5 Memory, Executive The Brief Memory and AD group performed worse on
(AD) functioning, Processing Executive Test semantic verbal fluency test
speed compared to HOA (P < 0.001).
HOA 80 (45F) 68.1 ± 7.9 General cognitive Mini-Mental State
impairment Examination (MMSE)
† Wechsler Memory Scale-Third Edition.
‡
controlled oral word association test.
subside and continuous progression of the disease occurs. in MS has provided conflicting results. Findings of decreased
A minority of patients (< 10%) have primary progressive CVR in MS patients compared to controls using transcranial
MS (PPMS) where the progressive form is evident from Doppler ultrasonography (TCD) [150] has been supported by
onset [142]. MRI indications of MS include multiple lesions reported global decreases in gray matter CVR using CBF-
dispersed through the white matter, in particular lesions sensitized arterial spin labeling (ASL) [151]. However, other
adjacent to the ventricles and cerebral cortex, as assessed studies also using TCD and ASL report no CVR differences
with T2 -weighted sequences. Lesions may also demonstrate between MS patients and controls [152, 153], leading to
gadolinium enhancement as evidenced on post-contrast T1 - an inconclusive picture regarding CVR impairment in MS.
weighted images, indicating acute localized breakdown of the However, variations in disease severity and cognitive impairment
blood brain barrier [143]. Even normal-appearing white matter amongst MS patients might underlie these conflicting CVR
(NAWM) on MRI may be affected by mild inflammation and results. One study reported no significant differences in CVR
gliosis in MS, despite these regions largely retaining myelin between MS patients and healthy control groups, but when
[144, 145]. While the pathogenesis of MS remains to be fully looking within the patient group only, CVR was significantly
elucidated, the most widely accepted proposition is that it is an reduced in those with cognitive impairment compared to
autoimmune disease, with inflammation and neurodegeneration those with intact cognition [154]. Studies pooling subjects in
playing a critical role in lesion development and disease terms of disease course (e.g., RRMS and SPMS) and cognitive
progression [146]. impairment may be a contributing factor to the inconsistent
Vascular dysfunction has been reported in patients with CVR findings.
MS [147]. Studies evaluating CBF have shown hypoperfusion Cognitive decline in MS can be found across all stages
in MS patients relative to healthy controls [148], including of disease progression. As demonstrated by Table 3 (which
regions with NAWM [149]. However, the literature on CVR followed the search criteria outlined in section Cerebral Small
Frontiers in Physics | www.frontiersin.org 6 April 2021 | Volume 9 | Article 645249Williams et al. CVR, Cognition and Task fMRI
damaged, communication between network nodes is less
efficient. However, communication affecting neuronal activity is
not the only breakdown, as the cells governing neurovascular
coupling are also damaged. Damage to the neurovascular unit
results in insufficient communication between neurons and the
vasculature, and unmet nutrient and oxygen requirements [147].
Reduced BOLD response magnitude to cognitive tasks in MS
patients may reflect decreased neural activity and/or disrupted
NVC, but disentangling these requires more information, such as
CVR mapping.
NORMAL PHYSIOLOGY
Caffeine
Caffeine is a commonly used ergogenic aid due to it being
a psychostimulant of the central nervous system. The most
prominent mechanism of action is as an adenosine receptor
A1 and A2A antagonist. By blocking the action of inhibitory
neurotransmitter adenosine on these receptors, caffeine produces
a stimulation effect through disinhibition [170]. It has been long
known that adenosine is a regulator of cerebral vasodilation
[171]. As an antagonist, caffeine causes a vasoconstrictive
effect and reduces CBF [172, 173]. Most studies investigating
the neurovascular effects of caffeine consumption administer
a dose of 200 mg, which decreases baseline CBF by 30–35%
and increases oxygen extraction fraction (OEF) by 15–27%
FIGURE 1 | (A) BOLD-CVR maps for a healthy control participant (left) and a
[174–177]. This inverse relationship between CBF and OEF is
pre-surgical patient with MMD (right). Negative CVR in the MMD patient can intuitive, as a reduction in flow would require an increase in
be observed anteriorly, while the posterior brain is well-preserved. (B) fMRI OEF to maintain oxygen metabolism rates. But does baseline
activation maps from a concurrent visual-motor task, showing robust visual CMRO2 change after caffeine consumption? The literature is
activation for the healthy control (left) and MMD patient (right). (C) Activation
inconsistent. For instance, baseline CMRO2 has been reported as
maps showing the motor cortex during the same task as (B). The healthy
control on the left shows preserved activation, while the motor cortex of the unchanged [177], increased [178] and decreased [176] following
MMD patient (right) shows NBRs due to vascular steal in blue. Further caffeine consumption.
information about the study and experimental protocol used to calculate these The effects of caffeine on CVR appears somewhat more
maps can be found in Mazerolle et al. [125]. Color bar indicates Z-values. consistent. When characterizing the influence of caffeine on CVR
in the motor and visual cortices, one study showed that caffeine
increased BOLD-CVR in these regions while having no effect
Vessel Disease), MS patients have demonstrated impairment on CBF-CVR [179]. Similarly, Merola et al. [176] reported an
across a broad range of cognitive domains. Cognitive decline is increase in BOLD-CVR across the gray matter after caffeine,
not ubiquitous; approximately half of all patients with MS do relative to caffeine-free baseline. Some example CVR maps from
not exhibit any impairment in cognition [163]. However, the individuals who participated in caffeine and caffeine-free placebo
presence of cognitive decline in MS has been associated with conditions are shown in Figure 2. This figure highlights some
reduced BOLD activation during cognitive tasks. Patients with increased CVR with caffeine, although inter-subject variability
poorer cognitive performance have shown decreased extent of is also evident. One interpretation that has been previously
BOLD activation to memory and attention tasks [164, 165]. MS linked to increased task-induced BOLD signal changes following
patients with intact task performance have shown additional caffeine is that a decreased CBF baseline may result in a larger
regions of brain activation compared to healthy controls [166– relative BOLD amplitude change [181]. However, further work
168]. This association between cognitive performance and BOLD found that caffeine-induced reductions in resting CBF were not
activation in MS has been explained as functional reorganization. a strong predictor of stimulus-induced BOLD activation [182].
In a study comparing MS progression sub-types, Loitfelder et al. Further research investigating how caffeine influences CVR, and
showed that the extent of BOLD activation to a processing speed whether this in dependent on factors such as baseline physiology,
task increased from patients with clinically isolated syndrome, is therefore warranted.
to RRMS, and was most extensive in patients with SPMS The cognitive effects of caffeine are well-known and
[169]. It was reasoned that this was evidence for neuroplasticity appreciated by frequent consumers. The ingestion of up to
changes to increasing tissue damage; more neural networks 300 mg has shown to enhance alertness, attention, and reaction
were required to compensate for those that were damaged by time performance [183, 184]. How caffeine affects task-induced
lesions and atrophy. As neural networks become increasingly BOLD signals appears to be dependent on the specific task
Frontiers in Physics | www.frontiersin.org 7 April 2021 | Volume 9 | Article 645249Williams et al. CVR, Cognition and Task fMRI
TABLE 2 | Summary of research studies highlighting the cognitive domains impaired in Moyamoya disease.
References Study population(s) N (N Mean age Cognitive domain(s) Tests administered Main findings
Female) assessed
He et al. [134] MMD with infarction 19 (9F) 41.9 ± 11.7 All cognitive domains Neuropsychology test MMD with infarction performed
battery worse than MMD asymptomatic
MMD asymptomatic 21 (7F) 39.4 ± 10.2 on tests of complex arithmetic
Healthy controls (HC) 20 (8F) 42.61 ± 3.9 (P = 0.03) and short-term
memory (P = 0.01).
Both MMD groups performed
worse than HC on tests of
intelligence, spatial imagination,
working memory and
computational ability (P < 0.02).
Shi et al. [135] MDD 49 (25F) 38.5 ± 9 Intelligence Wechsler adult intelligence MMD performed worse that HC
test on intelligence (P = 0.002).
HC 23 (14F) 37.6 ± 9.1 Prospective memory (PM) Cambridge prospective MMD performed worse that HC
memory test on PM (P < 0.0005).
Immediate memory; Repeatable battery for the MMD performed worse that HC
Retrospective memory assessment of the on tests of VF and RM
(RM); Verbal fluency (VF); neuropsychological status (P < 0.006).
Visual breadth test
Executive functioning Stroop test; Trail making MMD performed worse that HC
test; Wisconsin card sorting on tests of executive functioning
test; Continuous (P < 0.003).
performance test
Attention Posner attention test MMD performed worse that HC
on attention (P < 0.0005).
Miyoshi et al. MMD (over 2 years) 70 (57F) 43 ± 8 Intelligence Wechsler adult intelligence Unchanged cognitive scores
[136] scale- revised over 2 years.
Memory Wechsler memory scale
Visuospatial abilities Rey-osterreith complex
figure test
Fang et al. MMD 49 (25F) 27.9 ± 14.1 Executive functioning Stroop; Hayling Sentence MMD groups scored worse than
[137] Completion Test (HSCT); HC on VF, HSCT and SART
Verbal Fluency (VF); N-Back; (P < 0.001).
Sustained Attention to
Response Task (SART)
HC 47 (25F) 27.1 ± 14.7 Subjective feelings about Dysexecutive Questionnaire There was no difference between
executive functioning MDD and HC groups on the
Dysexecutive Questionnaire.
performed. There is some evidence showing no change to passive frontal pole to an attention task. An important feature of these
visual stimulation-induced BOLD activation following caffeine research findings is that BOLD modulations to cognitive tasks
consumption, despite significantly fewer activated CBF voxels induced by caffeine may not be replicated in populations with
acquired simultaneously [185]. In a different study [186] where cognitive decline. For instance, a longitudinal study evaluating
a visuo-motor task that required continual alertness to visual the effects of caffeine in older adults [189] reported that those
cues was implemented, significant BOLD increases with caffeine who exhibited cognitive decline across time were less sensitive
ingestion were found. These increases corresponded to regions to caffeine-induced BOLD modulations than those showing
that have been implicated in attention networks, including cognitive stability.
the basal ganglia, thalamus, putamen and insula. This finding The research discussed here highlights that caffeine modulates
is consistent with other work showing that caffeine increases both vascular and neural activity, and underscores how the
BOLD activation during a working memory task [187]. This BOLD signal is a complex interplay between these components.
discrepant effect of caffeine on BOLD activation to low-level As such, task-induced BOLD signal changes following caffeine
sensory tasks vs. high-level cognitive tasks was investigated in consumption are highly dependent on the neural networks
a single study [188]. This work showed that caffeine reduced engaged by the task, and the level of alertness it requires.
BOLD activation in the visual and motor regions to sensory tasks, BOLD-CVR appears to increase following caffeine ingestion;
and consistent with the prior findings outlined above, increased this suggests that changes in task-induced BOLD activation
activation in the superior frontal gyrus, paracingulate cortex, and with caffeine may be partly explained by vascular contributions.
Frontiers in Physics | www.frontiersin.org 8 April 2021 | Volume 9 | Article 645249Williams et al. CVR, Cognition and Task fMRI
TABLE 3 | Summary of research studies highlighting the cognitive domains impaired in multiple sclerosis.
References Study population(s) N (N Mean age Cognitive domain(s) Tests administered Main findings
Female) assessed
Goitia et al. RRMS 36 (30F) 39.2 ± 10.2 Fluid intelligence (FI) Matrix reasoning subtest of RRMS patients showed impaired
[155] wechsler adult intelligence scores on FI compared to HC
scale-III (P < 0.001).
HC 42 (29F) 37.1 ± 10.7 Executive functioning (EF) Wisconsin cart sorting test; RRMS patients showed impaired
Verbal fluency; Trail making scores on all tests of EF
test B compared to HC (P < 0.05).
Multitasking Hotel task There was no difference between
groups on hotel task.
Social cognition (SC) Faux pas RRMS patients showed impaired
scores on SC compared to HC
(P < 0.001).
Pitteri et al. RRMS 51 (38F) 36.7 ± 9.1 Information processing Videogame with RRMS patients were slower
[156] speed under low, manipulated visual-attention under low and medium load
HC 20 (7F) 34.2 ± 12.3 medium, and high load (P < 0.01) and less accurate
visual-attentional load under high load (P < 0.001) than
HC.
Carotenuto MS (RRMS, SPMS) 55 (32F) 45.9 ± 14.3 Olfactory function The University of MS patients showed lower
et al. [157] HC 20 (10F) 40.1 ± 13 Pennsylvania Smell olfactory function than HC (P =
Identification Test (UPSIT-40) 0.02).
Kouvatsou MS (RRMS, SPMS) 38 (23F) 36.9 ± 9.6 Working memory: Digit recall; Word recall; MS group scored worse than HC
et al. [158] Phonological loop (PL) Non-word recall on all measures of PL
(P < 0.0005).
HC 27 (24F) 33 ± 7.3 Working memory: Block recall; Mazes task; MS group scored worse than HC
Visuospatial sketchpad Visual patterns test on Block and Mazes Recall
(VS) (P < 0.0005).
Working memory: Central Backward digit recall; MS group scored worse than HC
executive (CE) Backward block recall; on all measures of CE
Listening recall (P < 0.0005).
Working memory: California verbal learning MS group scored worse than HC
Episodic buffer test (Greek adapted on Immediate story recall
version); Logical Memory I- (P < 0.0005).
immediate story recall
Aladro et al. RRMS 88 (58F) 41.6 ± 8 Episodic memory (EM) Logical memory I and II, and MS group scored worse on all
[159] HC 40 (24F) 41.6 ± 8.4 Family pictures I and II of the EM tests than HC (P < 0.003).
WMS-III†
Migliore et al. RRMS (no cognitive 24 (13F) 42.3 ± 9.4 Processing speed, Task-Switching paradigm RRMS patients showed worse
[160] impairment) Executive functioning performance than HC on
HC 25 (13F) 42.8 ± 10.2 task-switching (P < 0.05).
Planche et al. Late-RRMS (LRRMS, 41 (34F) 51.3 ± 10 All cognitive domains Neuropsychology test Information processing speed
[161] disease duration > battery was the most frequently impaired
10 years) domain across all groups.
SPMS 37 (28F) 50.2 ± 9.7 SPMS more frequently affected
than LRRMS in all domains
except language (P < 0.01).
PPMS 23 (14F) 52 ± 8.4 More SPMS than PPMS patients
impaired at visuospatial
construction (P < 0.05).
Roth et al. RRMS 20 (17F) 43.5 ± 8.9 Attention Attention Network Test ANT accuracy was lower for
[162] (ANT) SPMS than HC (P = 0.006).
SPMS 20 (16F) 50.7 ± 6.2 Information processing Simple Reaction Time Patients (both groups) had
speed (SRT); Choice Reaction longer reaction times than HC on
HC 40 (31F) 46.2 ± 8.7 Time (CRT) Tasks SRT and CRT (P < 0.001).
† wechsler memory scale- third edition.
Further work exploring the link between caffeine-induced Aging
changes to CVR and task-based BOLD activation maps Healthy aging of the brain includes structural and vascular
is required. changes [190]. Evidence for structural changes in healthy older
Frontiers in Physics | www.frontiersin.org 9 April 2021 | Volume 9 | Article 645249Williams et al. CVR, Cognition and Task fMRI
FIGURE 2 | Example BOLD-CVR maps from 4 participants (P1, P2, P3, and P4) who ingested placebo (upper rows) and caffeine (lower rows) pills across two
separate scan sessions, 48 h apart. The caffeine condition shows some increased CVR, such as in P1, although there is individual variation and increased CVR is not
immediately apparent in every individual. Negative CVR is also evident (in cool colors) in regions close proximity to the ventricles. CVR maps are overlaid on
participant’s own T1 -weighted structural image. Further information about the experimental protocol used to obtain these maps can be found in Specht et al. [180].
Color bar indicates CVR (%1BOLD/1mmHg).
adults includes cortical thinning, with frontal regions appearing reported that CVR decreased linearly with increasing age across
to show accelerated rates of thinning with increasing age relative the whole brain, but the temporal lobes showed the fastest rate of
to temporal and occipital regions [191, 192]. Vascular alterations decline. The occipital lobes were the most resistant to age-related
include changes in cerebral microvascular organization and decline in CVR. However, inter-subject variation was evident in
capillary function [193] and indicate a functional decline of this longitudinal study. Lifestyle factors are likely contributors to
the neurovascular unit with age [194, 195]. These alterations variation in CVR. Indeed, exercise and fitness level is positively
at the microvascular level may lead to the changes observed at associated with CVR in healthy older adults [204]. CVR can
the macrovascular level. Macrovascular aging is characterized also reflect cognitive changes with aging. Catchlove et al. [202]
by increased arterial stiffness, corresponding to increased showed that CVR in the temporal lobes was associated with
pulse pressure and impaired vascular endothelium function memory and attention performance in older adults. Further
[196], and increased prevalence of hypertension [197]. MRI investigations of healthy brain aging and CVR are required. In
demonstrates CBF decreases with age [190, 198]. CVR is a particular, it would be important to investigate how the temporal
useful method for tracking age-related changes to vascular characteristics of the CVR response relate to cognitive changes
health. Most studies utilizing CVR to investigate aging with aging.
have focused on age-related diseases; however, important Task-based fMRI studies have shown BOLD activation
research findings have elucidated CVR changes with disease- decreases with increasing age during tasks of visual attention
free aging to understand how the healthy brain progresses [205, 206], working memory [207, 208] and memory encoding
throughout the adult lifespan. In this section, it will be argued [209, 210], as would be expected if such cognitive functioning
that obtaining complementary physiological information decreases with age. However, neural activity-mediated BOLD
such as CVR should be considered a priority for all fMRI changes with age also show significant heterogeneity. One
investigations in aging populations, especially when comparing meta-analysis reported decreased prefrontal cortex activation in
to younger adults. older adults during working memory tasks, while other task-
CVR decreases in gray matter have been consistently reported engaged regions including the cingulate and parietal cortex
with healthy aging [198–200]. CVR changes with age have also remained unaffected throughout adulthood [207]. Other studies
shown spatial heterogeneity. In cross-sectional studies, decreases reveal conflicting findings by showing increased frontal lobe
in CVR within frontal [201] and temporal lobes [202] with activation in older adults compared to younger cohorts during
increasing age have been reported. To assess CVR changes over working memory tasks [211–213]. These inconsistent findings
time, a longitudinal study by Peng et al. [203] characterized reflect increased variability in fMRI activation patterns in the
changes over 4 years in participants aged 20–88 years. It was older groups.
Frontiers in Physics | www.frontiersin.org 10 April 2021 | Volume 9 | Article 645249Williams et al. CVR, Cognition and Task fMRI
Two example patterns of age-related changes to BOLD where subjects inhaled a 5% CO2 gas mixture, and performed a
activation that have been reported in the literature are motor task. The peak activation observed in the motor task prior
hemispheric asymmetry reduction, and posterior-to-anterior to normalization was in the sagittal sinus; after normalization
shift. The first describes a shift in BOLD activation to cognitive the peak signal shifted to the brain parenchyma and therefore
tasks from unilateral to bilateral with increasing age [214–216]. improved spatial specificity toward neuronal function.
That is, older adults more often recruit both cerebral hemispheres This work was followed up by Cohen et al. [222] who
for tasks that show unilateral engagement in younger adults. investigated the effects of magnetic field strength on hypercapnic
The recruitment of both hemispheres in older adults has been normalization. Similar to Bandettini and Wong [221], a motor
associated with better task performance, and has been suggested task and 5% CO2 hypercapnia stimulation was implemented.
to be a compensatory mechanism [217]. The second pattern, At a magnetic field strength of 4 T, the findings were consistent
the posterior-to-anterior shift, describes findings of decreased with Bandettini and Wong [221], where voxels containing large
activation in the occipital and temporal regions, and increase vessels were highly activated by the motor task but showed
frontal activation during a variety of cognitive tasks, and low normalized BOLD signal. Furthermore, these voxels were
this is also hypothesized to reflect compensatory mechanisms located in regions distant to the site of neural activity. At 7 T,
[218]. As outlined above, CVR also decreases with age, thus the effect of hypercapnic normalization was less apparent due
complicating the interpretation of task-based BOLD activation to the expected increased sensitivity to smaller blood vessels
shifts. This is illustrated in Figure 3. Gauthier et al. exemplified at this higher field strength. Research studies following Cohen
these interpretation challenges by demonstrating that groups et al. [222] capitalized on hypercapnic normalization to improve
of older and younger adults showed similar BOLD activations signal specificity. The breath-hold hypercapnia approach for
to the Stroop task (i.e., executive functioning), however, CVR increasing CBF was implemented to normalize BOLD-activation
was reduced in the older group [219]. This demonstrates that to a motor task [223] and a working memory task [224].
BOLD activations to cognitive tasks cannot be directly compared These studies found that hypercapnic normalization reduced the
between groups of different ages, as underling metabolic activity influence of vascular variability between subjects. A study by
cannot be inferred from BOLD activation maps alone. Similar Handwerker et al. [20] implemented hypercapnic normalization
BOLD activation profiles between the older and younger adults to reduce vascular variability in task-induced BOLD signal
in the context of reduced CVR means that metabolic activity was in order to improve group comparisons between older and
likely different between the groups [220]. Overall, the complex younger adults. Normalization removed age-related differences
neurobiological and heterogenous changes that occur with age in some of the regions of interest, in particular those in frontal
are reflected in the highly variable and inconsistent task-based brain regions. A strong linear relationship between task-induced
BOLD findings in the literature. and hypercapnia BOLD signal was found and the authors
suggested that hypercapnia response variability accounts for a
large amount of the BOLD signal variability observed during
INTEGRATION OF BOLD-BASED the task. However, inter-subject variation increased in the older
ASSESSMENTS OF CVR AND NEURAL group after hypercapnic normalization, in contrast to the findings
ACTIVITY: HYPERCAPNIC of Thomason et al. [224] and Biswal et al. [223].
NORMALIZATION These conflicting findings were examined in detail by Liau
et al. [225]. To address inter-subject variability, their study
This review paper has emphasized that adding CVR to task-based investigated different approaches to hypercapnic normalization.
fMRI protocols may improve interpretation of BOLD activation The first method utilized a voxel-wise approach where the
maps. One analytical approach that integrates CVR is using it to task-induced BOLD response for each voxel is divided by the
correct activation maps through hypercapnic normalization. This hypercapnic BOLD response from the corresponding voxel.
was initially suggested by Bandettini and Wong [221] as a way The second method was also a division tactic but used each
to minimize signal contributions from large blood vessels. The subject’s averaged hypercapnia BOLD responses within ROIs,
premise of this work was that BOLD signal magnitude is highly rather than a voxel-wise implementation. The selection of
weighted by baseline cerebral blood volume (CBV). They noted ROIs was based on voxels demonstrating significant task and
that a very large BOLD signal change of 15% would be observed hypercapnic BOLD activation. A third method shifted from
in a voxel with 20% CBV, compared to a moderate 2.5% signal the division tactic and modeled hypercapnic BOLD responses
change in a voxel with 5% CBV, assuming all other parameters are as covariates. The covariate-normalized approach fit the task-
identical. Removing this activation bias from voxels with higher induced BOLD response amplitudes to linear regression models
CBV might be achieved using hypercapnia-induced BOLD where the hypercapnic BOLD responses were regressors. The
signal changes. The authors suggested that hypercapnia induces amplitudes of the task-induced signals were then normalized
consistent venous oxygenation changes across the brain, and any by the slope of the regression lines. The authors found that
spatial variation in BOLD intensity would presumably be due to the covariate-normalized approach demonstrated reduced inter-
resting CBV and vessel size. The division of activation-induced subject variability compared to the division methods.
signal change by hypercapnia BOLD signal change would return a The inclusion of CVR information into task-based fMRI has
normalized task-induced activation map void of influences from shown to improve comparisons between older and younger
large vessels. This was investigated using a hypercapnia stimulus adults in a memory encoding task [226]. This study showed
Frontiers in Physics | www.frontiersin.org 11 April 2021 | Volume 9 | Article 645249You can also read
